Vehicle-to-Infrastructure Collaborative Spatial Perception via Multimodal Large Language Models

Kimia Ehsani; Walid Saad

REVIEW 4 major objections 5 minor 22 references

Vehicle-to-Infrastructure Collaborative Spatial Perception via Multimodal Large Language Models

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Injecting neighbor-vehicle bird's-eye-view tokens into a frozen multimodal LLM raises V2I link-prediction macro-accuracy from 73.3% to 87.2% in simulation.

desk verdict A coherent, novel idea for injecting collaborative BEV tokens into a frozen MLLM for V2I link prediction, but the headline zero-shot robustness claim is contradicted by the paper's own split and the evaluation is too closed-loop to support it as stated. read the letter →

arxiv 2509.03837 v1 pith:U2EZMMSU submitted 2025-09-04 cs.LG cs.ITmath.IT

Kimia Ehsani , Walid Saad This is my paper

classification cs.LGcs.ITmath.IT

keywords vehicle-to-infrastructure(V2I)multimodallargelanguagemodelsbird's-eyeviewfusioncollaborativesensinglinkpredictionmmWaveraytracingblockagezero-shotgeneralization

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to show that an off-the-shelf multimodal LLM can reason about three-dimensional space for wireless link quality if it is given compact bird's-eye-view (BEV) tokens built from other vehicles' sensors. A lightweight trainable connector fuses each helper vehicle's RGB and LiDAR into a BEV, aligns those maps into the ego vehicle's frame, distills the task-relevant spatial cues, and injects them as tokens into a frozen LLM. Across LoS/NLoS classification, link availability, and blockage risk prediction on a CARLA-plus-ray-tracing dataset, this lifts macro-average accuracy by 13.9 percentage points over an ego-only LLM baseline, and the margin grows to 32.7 points under zero-shot rainy and nighttime conditions. If true, this would let one frozen LLM serve multiple V2I link-prediction tasks with only a small connector trained, rather than retraining a large model for each task.

What carries the argument

The carrying object is the BEV-injection connector: a trainable, architecture-agnostic module inserted between frozen vision encoders and a frozen LLM. It chains BEVFusion (per-agent RGB+LiDAR into a bird's-eye view), BEVFormer temporal self-attention (motion cues over a short frame window), a GPS-based warp into the ego frame, a 3x3 convolution for multi-agent aggregation, and an instruction-aware Q-Former that distills the aggregated BEV into compact spatial tokens. Those tokens are fused with the ego BEV via cross-attention and fed to the LLM alongside the text instruction. Only the connector's parameters are trained; the vision encoder and LLM remain fixed. The co-simulation pipeline (CA

What would settle it

Re-run the same three tasks with 1-3 m Gaussian noise added to helper-vehicle poses before the GPS warp; if the macro-average accuracy advantage over the ego-only LLM baseline collapses from 13.9 points toward zero, the collaborative gain depends on perfect pose alignment, not genuine spatial perception.

Watch

Extended reading notes

Core claim

The central claim is that spatial perception for V2I link prediction does not require retraining a large model; it can be added through a trainable connector that turns collaborative sensor data into a shared bird's-eye view and injects distilled tokens into a frozen MLLM. Concretely, BEVFusion combines each vehicle's multi-view RGB and LiDAR into a local BEV; BEVFormer adds temporal motion context; the helper BEVs are warped into the ego frame using GPS and aggregated; and an instruction-aware Q-Former compresses the aggregated map into a few tokens the LLM can attend to. On the paper's simulated dataset, this design achieves 87.2% macro-average accuracy across the three link-prediction tas

Load-bearing premise

All accuracy numbers are computed against ground truth produced by a MATLAB ray tracer running on a converted CARLA mesh, and the warp step assumes helper-vehicle GPS poses are exact; if the ray tracer does not faithfully model real mmWave V2I propagation or pose errors are significant, the reported gains may not transfer to real roads.

Editorial extensions

If this is right

One frozen LLM backbone, with only the connector trained, can handle three different V2I link-prediction tasks at once instead of requiring a dedicated model per task.
The largest gains are in LoS/NLoS and blockage prediction, the two tasks that need geometric context beyond the ego view, so the mechanism works by filling blind spots.
Most of the spatial benefit comes from the first one or two helper vehicles; additional helpers add little accuracy.
The connector retains most of its accuracy when tested zero-shot in rain and night, suggesting that geometry-based BEV tokens are less sensitive to visual appearance changes than raw image features.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The paper never corrupts the helper-vehicle poses in the warp step; adding realistic 1-3 m GPS noise in simulation would show how much of the 13.9-point margin depends on perfect alignment rather than on robust spatial reasoning.
Because the connector is trainable-only and backbone-agnostic, the same injection pattern could plausibly be carried over to other spatial communication tasks such as beam selection, handover timing, or resource allocation without retraining the LLM.
The 32.7-point zero-shot advantage is measured against ray-traced ground truth; a real urban measurement campaign with synchronized RGB, LiDAR, GPS, and received power would be the decisive test of whether that advantage survives physical propagation.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The idea here is genuinely new: take a frozen vision-language model, feed it a compact BEV built from multiple cooperative vehicles, and show it beats single-agent baselines on three link-quality tasks. That combination—multi-agent BEV fusion plus MLLM—doesn't appear in the cited LLM-for-wireless papers, and the architecture is sensible. The plug-and-play framing is also reasonable; keeping the LLM and encoder frozen while training only a connector is the right way to test the hypothesis. The paper does a service by building a CARLA+ray-tracing co-simulation and releasing nothing, which is a problem I'll get to.

Where it falls short is the evidence for the strongest claim. The abstract and Section IV.D advertise a 32.7-percentage-point gain under 'zero-shot' rain and night. But Section IV.B says the full 50-episode set is split 80/10/10 randomly over all episodes, which means some rain and night episodes are almost certainly in the training set. The rain/night bars in Figure 4 are then in-distribution, not zero-shot. If the split held out all 20 adverse episodes, the test set would be about 5 episodes and the 32.7-point gap would be a single-run point estimate with no error bars. Either way, the robustness claim is unverifiable as written. That's not a nitpick; it's the headline result.

There are two more soft spots, both real but less severe. The ego-only LLM baseline has no trainable connector, so the daytime 13.9-point gain conflates added adapter capacity with BEV spatial content. Table II ablates components but never holds trainable capacity fixed, so you can't isolate the information contribution of the multi-agent BEV. Second, the entire evaluation is a closed loop: labels are derived programmatically from the same ray tracer that generates the inputs, and there's no external benchmark or real channel data. That makes every accuracy number a measure of the simulator's fidelity, and the simulator's fidelity to real mmWave propagation is never validated. These two problems compound rather than cancel.

What's solid: the architecture is coherent, the math is fine, and the controlled comparisons (when the split is properly stated) do favor the proposed method. The authors are not sloppy in the mechanics; they just overclaim on the zero-shot front and underreport uncertainty. The paper deserves a serious referee—the idea is worth engaging—but it needs major revision before the robustness claim can be trusted. Concretely: re-run the zero-shot evaluation with a clean split, add error bars, add the missing connector ablation, and release the code and data. I'd cite the idea once it's clean, but not in its current form.

Referee Report

4 major / 5 minor

Summary. The paper proposes a plug-and-play bird's-eye-view (BEV) injection connector that fuses LiDAR and RGB data from the ego vehicle and cooperative helper vehicles into compact spatial tokens, which are injected into a frozen multimodal large language model (Llama-3.2-11B-Vision). The connector consists of BEV fusion, temporal attention, warping to the ego frame, and an instruction-aware Q-Former. The authors also build a custom dataset by co-simulating CARLA with MATLAB ray tracing, covering 50 urban episodes and three V2I link prediction tasks: LoS/NLoS classification, link availability, and blockage risk prediction. The main claims are that BEV injection improves macro-average accuracy from 73.3% (ego-only LLM baseline) to 87.2%, and that the gain grows to 32.7 percentage points under zero-shot transfer to rainy and nighttime conditions. The central quantitative claims are currently not fully supported by the stated evaluation protocol.

Significance. If validated, the idea of giving a frozen LLM 3D spatial reasoning through a lightweight trainable BEV connector is interesting and relevant to V2I beam management, blockage prediction, and cooperative perception. The modular architecture is a strength, as is the purpose-built simulation dataset with programmatically extracted ground truth. However, the zero-shot robustness claim is undermined by an ambiguous train/test split, and the main comparison is confounded by unequal trainable capacity between the proposed method and the ego-only baseline. Confidence intervals are absent, and the closed-loop simulation evaluation limits the strength of the physical conclusions. The framework is plausible, but the evidence presented does not yet support the advertised margins.

major comments (4)

[IV.B/IV.D, Fig. 4] The zero-shot claim is not supported by the stated data split. Section IV.A states that the dataset has 30 noon, 10 night, and 10 rain episodes, while Section IV.B says the dataset is partitioned 80/10/10 into training, validation, and test without conditioning on scenario. If the split is random over all episodes, then rainy and nighttime episodes appear in the training set, making Fig. 4 an in-distribution evaluation rather than zero-shot transfer. If the authors intended to train only on noon episodes, the text never says so, and an 80/10/10 split cannot hold out all 20 adverse-condition episodes while preserving a 10% test set. Furthermore, no confidence intervals are reported, so the 32.7-point margin is a point estimate from a very small test set. Please either retrain with an explicit clear-day training split and report held-out adverse-condition results with error bars, or revise
[Table I and Section III] The main comparison is confounded by unequal trainable capacity. The ego-only LLM baseline has no trainable connector, while the proposed model adds a trainable BEV-injection connector with a Q-Former. The 13.9-point macro-average gain (73.3 vs. 87.2) may therefore reflect the added adapter capacity rather than the multi-agent BEV content. Table II removes components from the full BEV-injection model, but no condition applies the same trainable connector to ego-only features or keeps the connector capacity fixed while varying the input BEV content. Add an ablation with (a) ego-only features passed through the same Q-Former/connector and (b) multi-agent BEV tokens replaced by an ego-only BEV token set of identical size.
[IV.A and II.A] The evaluation is a closed loop: ground-truth labels for all three tasks are programmatically extracted from the ray-tracing outputs (Section IV.A), and the ray tracer is also the source of the channel model in Section II.A. Consequently, the reported accuracies measure consistency with the simulator, not agreement with physical V2I channels. External validation against measured mmWave data or a public benchmark, or at least an explicit and quantified statement of the simulator's fidelity, is needed before the robustness claims can support real-world conclusions.
[Eq. (9), Section I.B] The multi-agent BEV warp assumes exact relative poses, described as alignment through 'precise coordinate-frame transformation.' Real V2I cooperative perception is subject to GPS and localization errors. The reported gains may rely on perfect pose information and may not transfer to realistic deployments. Add a sensitivity analysis that injects pose noise into Eq. (9) and reports how the macro-average accuracy degrades.

minor comments (5)

[Abstract and Section V] The text uses '%' where it means percentage points. For example, 73.3% to 87.2% is a 13.9 percentage-point gain, not a 13.9% relative gain. Please use consistent terminology throughout.
[Throughout] Typographical issues: 'Index T erms', 'he work in [11]', 'compared conventional approaches' (missing 'to'), and 'the architecture and of our' should be corrected.
[IV.B] The paper says the dataset is split 80/10/10, but it does not state whether the split is by episode or by frame. This is essential for interpreting Fig. 4 and for reproducing the evaluation; please clarify.
[Tables I and II] No standard deviations or confidence intervals are reported. Given the small number of episodes, especially under adverse conditions, report per-episode variance or bootstrap intervals to assess whether the observed margins are statistically meaningful.
[Figure 4] The legend and axis title ('Task Accuracy vs. Ego-only Baseline Across Scenarios (Zero-Shot)') are unclear: it appears to show both the BEV-injection model and the baseline, but the title suggests a comparison plot. Make the curves/bars explicit and mention the number of test episodes per scenario.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline zero-shot robustness gain reduces to in-distribution fitting under the paper's stated 80/10/10 split.

fitted input called prediction [Section IV.A, IV.B, IV.D and Figure 4]
"To increase diversity, 30 episodes occur at noon, 10 at night, and 10 in rain. ... Our dataset was partitioned into training (80%), validation (10%), and test (10%) subsets. ... Figure 4 shows zero-shot generalization from the clear daytime training set to unseen rainy and nighttime conditions."

The dataset has 50 episodes, of which only 30 are noon. An 80/10/10 split gives 40 training episodes. If the training set were 'clear daytime' only, it could contain at most the 30 noon episodes, i.e., 60%, not 80%. Therefore, under the stated split, at least 10 rainy/night episodes must fall in the training set. Hence the rainy/night test performance in Figure 4 cannot be 'unseen'; the connector was fitted on those conditions. The claimed 32.7-percentage-point zero-shot margin is thus in-distribution accuracy on fitted inputs, not an out-of-sample prediction.

full rationale

The paper's central advertised result is the 32.7-point robustness gain under rainy and nighttime conditions, presented as zero-shot generalization. This claim is contradicted by the paper's own split arithmetic: with 50 episodes and only 30 noon episodes, an 80% training split necessarily includes at least 10 rainy/night episodes. The 'clear daytime training set' and 'unseen rainy and nighttime conditions' cannot both hold, so the reported margin is at least partly in-distribution performance. This is a specific, non-speculative reduction. Other potential concerns are noted but not scored as circularity: ground-truth labels are programmatically extracted from the same ray tracer that generates the data, which is standard for simulation studies and only limits external validity; self-citations to Saad et al. are not load-bearing; and the ego-only baseline lacks a trainable connector, confounding adapter capacity with BEV content, but this is an ablation control issue rather than a circular derivation. The daytime 13.9% gain and the ablations retain independent content, so the overall circularity is partial but not total.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The central claim is empirical and rests on design choices whose sensitivity is unreported (5 helper vehicles, 3-frame windows, -80 dBm threshold, 80/10/10 split) plus unvalidated simulator realism. The ledger records the hand-chosen parameters, the domain assumptions about simulation fidelity and pose accuracy, and the one new architectural entity (the connector) whose evidence is only the authors' own closed simulation.

free parameters (5)

Link availability threshold = -80 dBm
Defines the ground-truth label for Task 2; chosen to ensure QPSK over 100 MHz, so it is a justified domain choice rather than a fitted constant, but it directly determines the reported availability accuracies.
Number of helper vehicles = 5
Fixed at 5 in every scenario; Figure 3 shows accuracy varies with agent count, so the reported numbers depend on this hand-chosen value.
Temporal window (key frames per episode) = 3 sampled frames
Section IV.B: 'three key frames were sampled in sequence'; the blockage-prediction task and the BEVFormer temporal self-attention depend on this hand-chosen window.
Training hyperparameters = lr 1e-4, batch 16, 15 epochs, weight decay 0.05, 5% warmup
Standard choices but hand-picked; no sensitivity analysis is reported.
BEV grid resolution and Q-Former token count = Not stated
The paper never specifies the BEV tensor size or the number of distilled tokens, both of which control the information bottleneck and the compute budget.

assumptions (4)

domain assumption The MATLAB ray tracer in Eqs. (1)-(2) produces correct ground-truth labels for LoS, availability, and blockage.
All three tasks' labels are 'programmatically extracted from the ray-tracing outputs' (Section IV.A); there is no validation against measured channels or an independent channel model.
domain assumption CARLA Town 10 renderings and the noon/rain/night splits are representative of real urban V2I conditions.
The transfer claim of the whole paper ('robustness in adverse settings') assumes simulator realism; no real-world data is used anywhere.
domain assumption GPS poses are noiseless, making the helper-to-ego warp in Eq. (9) exact.
'Aligned through precise coordinate-frame transformation' (Section I.B); no pose-error model is tested despite GPS noise being a known V2I issue.
domain assumption A frozen Llama-3.2-11B-Vision backbone provides sufficient vision-language reasoning for all three tasks.
Tooling assumption stated in Section IV.B; the paper does not test alternative backbones or the effect of unfreezing.

invented entities (1)

BEV-injection connector (trainable fusion plus Q-Former module)
purpose: Aggregates multi-agent BEV maps and injects compact spatial tokens into a frozen LLM so it can reason about 3D geometry for link prediction
This is a new architectural component, not a physical entity, but it is the paper's key introduced object; its only evidence is the authors' own closed simulation, so it has no independent falsifiable handle outside the paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Vehicle-to-Infrastructure Collaborative Spatial Perception via Multimodal Large Language Models." pith.science (2026). https://pith.science/paper/U2EZMMSU

@misc{pith2026250903837,
  author       = {Pith},
  title        = {Pith review of: Vehicle-to-Infrastructure Collaborative Spatial Perception via Multimodal Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U2EZMMSU}},
  note         = {Machine review of arXiv:2509.03837}
}

read the original abstract

Accurate prediction of communication link quality metrics is essential for vehicle-to-infrastructure (V2I) systems, enabling smooth handovers, efficient beam management, and reliable low-latency communication. The increasing availability of sensor data from modern vehicles motivates the use of multimodal large language models (MLLMs) because of their adaptability across tasks and reasoning capabilities. However, MLLMs inherently lack three-dimensional spatial understanding. To overcome this limitation, a lightweight, plug-and-play bird's-eye view (BEV) injection connector is proposed. In this framework, a BEV of the environment is constructed by collecting sensing data from neighboring vehicles. This BEV representation is then fused with the ego vehicle's input to provide spatial context for the large language model. To support realistic multimodal learning, a co-simulation environment combining CARLA simulator and MATLAB-based ray tracing is developed to generate RGB, LiDAR, GPS, and wireless signal data across varied scenarios. Instructions and ground-truth responses are programmatically extracted from the ray-tracing outputs. Extensive experiments are conducted across three V2I link prediction tasks: line-of-sight (LoS) versus non-line-of-sight (NLoS) classification, link availability, and blockage prediction. Simulation results show that the proposed BEV injection framework consistently improved performance across all tasks. The results indicate that, compared to an ego-only baseline, the proposed approach improves the macro-average of the accuracy metrics by up to 13.9%. The results also show that this performance gain increases by up to 32.7% under challenging rainy and nighttime conditions, confirming the robustness of the framework in adverse settings.

Figures

Figures reproduced from arXiv: 2509.03837 by the authors.

**Figure 1.** Overall architecture performs collaborative BEV fusion in the input stream of a frozen LLM to encourage spatial understanding. LLMs, while powerful at language reasoning, lack explicit 3D spatial priors and cannot “see around corners.” To bridge this gap, we introduce a plug-and-play BEV-injection connector that (i) preserves all frozen vision encoders and the LLM intact, (ii) uses underexploited sensing data from … view at source ↗

**Figure 2.** Data generation using CARLA with MATLAB mmWave ray tracing. this end, we apply a BEVFormer [15] temporal self attention (TSA) mechanism over the fused BEV sequence to produce a motion-aware BEV feature for vehicle 𝑣𝑖 : 𝑩 local 𝑖 = TSA n BEVFusion 𝑩 (𝜏) img, 𝑩 (𝜏) lidaro𝑡+Δ 𝜏=𝑡−Δ (8) Since each BEV map is constructed in the local coordinate frame of its agent, we warp it into the ego vehicle’s frame using relat… view at source ↗

**Figure 3.** shows the accuracy of our V2I link-quality prediction tasks as a function of the number of helper agents, where a helper agent is a neighboring vehicle that shares its local sensor data for cooperative BEV construction. In this figure, we observe that one or two neighbors rapidly improve accuracy by covering blind spots, while additional agents give smaller gains. This indicates that even a couple of well-positioned… view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: shows zero-shot generalization from the clear daytime training set to unseen rainy and nighttime conditions. While achieving impressive daytime performance, the BEV-injection model truly distinguishes itself under 0 1 2 3 4 5 Number of Helper Agents 50 60 70 80 90 100…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references · 10 canonical work pages

[1]

Artificial General Intelligence (AGI)- Native Wireless Systems: A Journey Beyond 6G,

W. Saad, O. Hashash, C. K. Thomas, C. Chaccour, M. Debbah, N. Mandayam, and Z. Han, “Artificial General Intelligence (AGI)- Native Wireless Systems: A Journey Beyond 6G,”Proceedings of the IEEE, pp. 1–39, 2025

work page 2025
[2]

Joint Sensing, Communication, and AI: A Trifecta for Resilient THz User Experiences,

C. Chaccour, W. Saad, M. Debbah, and H. V. Poor, “Joint Sensing, Communication, and AI: A Trifecta for Resilient THz User Experiences,”IEEE Transactions on Wireless Communications, vol. 23, no. 9, pp. 11444–11460, Sep. 2024

work page 2024
[3]

Sensing-Assisted High Reliable Communication: A Transformer- Based Beamforming Approach,

Y. Cui, J. Nie, X. Cao, T. Yu, J. Zou, J. Mu, and X. Jing, “Sensing-Assisted High Reliable Communication: A Transformer- Based Beamforming Approach,”IEEE Journal of Selected Topics in Signal Processing, vol. 18, no. 5, pp. 782–795, Jul. 2024

work page 2024
[4]

Y. Tian, Q. Zhao, Z. e. a. Kherroubi, F. Boukhalfa, K. Wu, and F. Bader, “Multimodal Transformers for Wireless Communications: A Case Study in Beam Prediction,arXiv:2309.11811 [eess], Sep. 2023

arXiv 2023
[5]

Vision-Aided 6G Wireless Communications: Blockage Prediction and Proactive Handoff,

G. Charan, M. Alrabeiah, and A. Alkhateeb, “Vision-Aided 6G Wireless Communications: Blockage Prediction and Proactive Handoff,” IEEE Transactions on Vehicular Technology, vol. 70, no. 10, pp. 10193–10208, Oct. 2021

work page 2021
[6]

Passive Radar at the Roadside Unit to Configure Millimeter Wave Vehicle-to-Infrastructure Links,

A. Ali, N. González-Prelcic, and A. Ghosh, “Passive Radar at the Roadside Unit to Configure Millimeter Wave Vehicle-to-Infrastructure Links,” IEEE Transactions on Vehicular Technology, vol. 69, no. 12, pp. 14903–14917, Dec. 2020

work page 2020
[7]

Zheng, J

C. Zheng, J. He, G. Cai, Z. Yu, and C. G. Kang, “BeamLLM: Vision-Empowered mmWave Beam Prediction with Large Language Models, arXiv:2503.10432 [cs], Mar. 2025

arXiv 2025
[8]

LLM4CP: Adapting Large Language Models for Channel Prediction,

B. Liu, X. Liu, S. Gao, X. Cheng, and L. Yang, “LLM4CP: Adapting Large Language Models for Channel Prediction,”Journal of Communications and Information Networks, vol. 9, no. 2, pp. 113–125, Jun. 2024

work page 2024

Show all 22 references

[9]

Zhang, H

Y. Zhang, H. Yin, W. Li, E. Bjornson, and M. Debbah, “Port-LLM: A Port Prediction Method for Fluid Antenna based on Large Language Models, arXiv:2502.09857 [eess], Feb. 2025

2025 arXiv
[10]

Visual instruction tuning

H. Liu, C. Li, Q. Wu, and Y. J. Lee, “Visual instruction tuning.” [Online]. Available: https://arxiv.org/abs/2304.08485 2023

2023 arXiv
[11]

Large Language Models Empower Multimodal Integrated Sensing and Communication,

L. Cheng, H. Zhang, B. Di, D. Niyato, and L. Song, “Large Language Models Empower Multimodal Integrated Sensing and Communication,” IEEE Communications Magazine, vol. 63, no. 5, pp. 190–197, May 2025

2025
[12]

D. Yu, R. Bao, G. Mai, and L. Zhao, “Spatial-RAG: Spatial Retrieval Augmented Generation for Real-World Spatial Reasoning Questions, arXiv:2502.18470 [cs] version: 2, Feb. 2025

2025 arXiv
[13]

W. Dai, J. Li, D. Li, A. M. H. Tiong, J. Zhao, W. Wang, B. Li, P. Fung, and S. Hoi, “InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning,arXiv:2305.06500, Jun. 2023

2023 arXiv
[14]

BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird’s-Eye View Representation,

Z. Liu, H. Tang, A. Amini, X. Yang, H. Mao, D. L. Rus, and S. Han, “BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird’s-Eye View Representation,” in2023 IEEE International Conference on Robotics and Automation (ICRA), May 2023, pp. 2774–2781

2023
[15]

Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Q. Yu, and J. Dai, “BEVFormer: Learning Bird’s-Eye-View Representation from Multi- Camera Images via Spatiotemporal Transformers,arXiv:2203.17270 [cs], Jul. 2022

2022 arXiv
[16]

Dosovitskiy, G

A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V. Koltun, “CARLA: An Open Urban Driving Simulator,arXiv:1711.03938 [cs], Nov. 2017

2017 arXiv
[17]

Y. M. Park, Y. K. Tun, W. Saad, and C. S. Hong, “Resource-Efficient Beam Prediction in mmWave Communications with Multimodal Realistic Simulation Framework,arXiv:2504.05187 [cs], Apr. 2025

2025
[18]

Llama 3.2: Revolutionizing edge AI and vision with open, customiz- able models

“Llama 3.2: Revolutionizing edge AI and vision with open, customiz- able models.” [Online]. Available: https://ai.meta.com/blog/llama-3-2- connect-2024-vision-edge-mobile-devices/

2024
[19]

Decoupled weight decay regularization

I. Loshchilov and F. Hutter, “Decoupled weight decay regularization.” [Online]. Available: https://arxiv.org/abs/1711.05101 2019

2019 arXiv
[20]

Long Short-Term Memory,

S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, Nov. 1997

1997
[21]

Chung, C

J. Chung, C. Gulcehre, K. Cho, and Y. Bengio, “Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling, arXiv:1412.3555 [cs], Dec. 2014

2014 arXiv
[22]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention Is All You Need, arXiv:1706.03762 [cs], Aug. 2023

2023 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

Artificial General Intelligence (AGI)- Native Wireless Systems: A Journey Beyond 6G,

W. Saad, O. Hashash, C. K. Thomas, C. Chaccour, M. Debbah, N. Mandayam, and Z. Han, “Artificial General Intelligence (AGI)- Native Wireless Systems: A Journey Beyond 6G,”Proceedings of the IEEE, pp. 1–39, 2025

work page 2025

[2] [2]

Joint Sensing, Communication, and AI: A Trifecta for Resilient THz User Experiences,

C. Chaccour, W. Saad, M. Debbah, and H. V. Poor, “Joint Sensing, Communication, and AI: A Trifecta for Resilient THz User Experiences,”IEEE Transactions on Wireless Communications, vol. 23, no. 9, pp. 11444–11460, Sep. 2024

work page 2024

[3] [3]

Sensing-Assisted High Reliable Communication: A Transformer- Based Beamforming Approach,

Y. Cui, J. Nie, X. Cao, T. Yu, J. Zou, J. Mu, and X. Jing, “Sensing-Assisted High Reliable Communication: A Transformer- Based Beamforming Approach,”IEEE Journal of Selected Topics in Signal Processing, vol. 18, no. 5, pp. 782–795, Jul. 2024

work page 2024

[4] [4]

Y. Tian, Q. Zhao, Z. e. a. Kherroubi, F. Boukhalfa, K. Wu, and F. Bader, “Multimodal Transformers for Wireless Communications: A Case Study in Beam Prediction,arXiv:2309.11811 [eess], Sep. 2023

arXiv 2023

[5] [5]

Vision-Aided 6G Wireless Communications: Blockage Prediction and Proactive Handoff,

G. Charan, M. Alrabeiah, and A. Alkhateeb, “Vision-Aided 6G Wireless Communications: Blockage Prediction and Proactive Handoff,” IEEE Transactions on Vehicular Technology, vol. 70, no. 10, pp. 10193–10208, Oct. 2021

work page 2021

[6] [6]

Passive Radar at the Roadside Unit to Configure Millimeter Wave Vehicle-to-Infrastructure Links,

A. Ali, N. González-Prelcic, and A. Ghosh, “Passive Radar at the Roadside Unit to Configure Millimeter Wave Vehicle-to-Infrastructure Links,” IEEE Transactions on Vehicular Technology, vol. 69, no. 12, pp. 14903–14917, Dec. 2020

work page 2020

[7] [7]

Zheng, J

C. Zheng, J. He, G. Cai, Z. Yu, and C. G. Kang, “BeamLLM: Vision-Empowered mmWave Beam Prediction with Large Language Models, arXiv:2503.10432 [cs], Mar. 2025

arXiv 2025

[8] [8]

LLM4CP: Adapting Large Language Models for Channel Prediction,

B. Liu, X. Liu, S. Gao, X. Cheng, and L. Yang, “LLM4CP: Adapting Large Language Models for Channel Prediction,”Journal of Communications and Information Networks, vol. 9, no. 2, pp. 113–125, Jun. 2024

work page 2024